Method and device for labeling equipment knowledge graph data

Abstract

The embodiment of the invention provides an equipment knowledge graph data labeling method and device, and belongs to the technical field of industrial and mining equipment. The method comprises the following steps: collecting equipment characteristic data of a plurality of pieces of industrial and mining equipment, and performing data preprocessing on the equipment characteristic data to generate a data source which is de-noised and is uniform in data format; a labeling template is obtained, the labeling template and the data source are input into a pre-training large language model for preliminary labeling processing, preliminary labeling data corresponding to the multiple pieces of industrial and mining equipment are generated, and the preliminary labeling data comprise entities, entity types and entity relations corresponding to the multiple pieces of industrial and mining equipment; and performing semi-automatic evaluation processing on the entities, the entity types and the entity relationships corresponding to the plurality of pieces of industrial and mining equipment based on the data source, and generating final annotation data. The problems of high labor cost and time cost and low labeling accuracy caused by a labeling method in the prior art are solved.

Description

Technical Field

[0001] The present application relates to the field of mining equipment technology, specifically to the field of industrial and mining equipment data processing technology, and in particular to a method and device for labeling equipment knowledge graph data. Background Art

[0002] With the development of industrial Internet of Things and big data technology, efficient operation and maintenance and intelligent management of industrial and mining equipment have gradually become key research issues in the industry. In order to achieve efficient operation and maintenance and intelligent management of industrial and mining equipment, it is necessary to label the characteristic data of industrial and mining equipment to improve data utilization.

[0003] At present, in the existing technology, the labeling of massive data of industrial and mining equipment mainly relies on manual labor. However, due to the reliance on manual labeling, accurate labeling of a large amount of industrial and mining equipment data not only requires a lot of labor costs and time costs, but also leads to labeling errors due to the prolonged labeling time, manual subjectivity and inconsistency, resulting in low labeling accuracy.

[0004] Therefore, there is an urgent need for a labeling method for equipment knowledge graph data to improve labeling efficiency and accuracy, save labor costs and time costs, and provide more accurate data for building equipment knowledge graphs for industrial and mining equipment, so as to facilitate the subsequent efficient operation and maintenance and intelligent management of industrial and mining equipment. Summary of the invention

[0005] The embodiments of the present application provide a method and apparatus for labeling device knowledge graph data, which solve the problems of high labor and time costs and low labeling accuracy caused by the labeling methods in the prior art.

[0006] In a first aspect, the present application provides a method for labeling device knowledge graph data, including:

[0007] Collecting equipment characteristic data of multiple industrial and mining equipment, and performing data preprocessing on the equipment characteristic data to generate a data source with noise removed and a unified data format;

[0008] Obtaining a labeling template, and inputting the labeling template and the data source into a pre-trained large language model for preliminary labeling processing, to generate preliminary labeling data corresponding to each of the plurality of industrial and mining equipment, wherein the preliminary labeling data includes entities, entity types, and entity relationships corresponding to the plurality of industrial and mining equipment;

[0009] Based on the data source, semi-automatic evaluation and processing are performed on the entities, entity types and entity relationships corresponding to the plurality of industrial and mining equipment to generate final labeled data.

[0010] In a possible implementation, the data preprocessing of the device characteristic data to generate a data source with noise removed and a unified data format includes: cleaning the device characteristic data to generate noise-removed data; and formatting the noise-removed data to generate data in a unified format as source data.

[0011] In a possible implementation, the obtaining of the annotation template includes: defining a classification standard and different first entity categories for each type of industrial and mining equipment based on pre-stored equipment characteristic data; defining a first attribute field for each first entity category and determining a first relationship type between the first entity categories;

[0012] A labeling template is determined according to the first entity category, the first attribute field and the relationship type.

[0013] In a possible implementation, the inputting the data source into a pre-trained large language model to generate preliminary annotation data corresponding to each of the plurality of industrial and mining equipment includes: inputting the data source into a pre-trained large language model so that the pre-trained large language model automatically identifies key equipment information from the data source, classifies the key equipment information to obtain a plurality of entities and an entity type corresponding to each entity, and performs relationship extraction on the data source to obtain entity relationships between different entities.

[0014] In a possible implementation, the obtaining of the annotation template also includes: periodically obtaining annotation requirements, and determining a second entity category based on the annotation requirements; defining a second attribute field for the second entity category and determining a second relationship type between each second entity category; and determining a new annotation template based on the first entity category, the first attribute field, the first relationship type, the second entity type, the second attribute field, and the second relationship category.

[0015] In a possible implementation, it also includes: obtaining original industrial and mining equipment data, and performing preprocessing operations on the original industrial and mining equipment data to obtain a training sample set and a verification sample set, wherein the training sample set includes multiple first industrial and mining equipment characteristic data, and the verification sample set includes multiple second industrial and mining equipment characteristic data; according to the first industrial and mining equipment characteristic data, the initial large language model is trained to obtain the large language model to be adjusted; through multiple second equipment characteristic data, the labeling accuracy of the large language model to be adjusted is determined, if the labeling accuracy is less than a first preset threshold, the first adjustment operation is performed on the large language model to be adjusted until the labeling accuracy is greater than or equal to the first preset threshold, so as to obtain a pre-trained large language model.

[0016] In a possible implementation, after obtaining the original industrial and mining equipment data, the method further includes: performing data augmentation processing on the original industrial and mining equipment data through the initial large language model to obtain augmented industrial and mining equipment data, wherein the augmented industrial and mining equipment data includes the original industrial and mining equipment data and newly added industrial and mining equipment data.

[0017] In a possible implementation, the method further includes: determining the labeling accuracy of the pre-trained large language model through the final labeling data; if the labeling accuracy is less than a second preset threshold, performing a second adjustment operation on the pre-trained large language model until the labeling accuracy is greater than or equal to the second preset threshold, thereby obtaining a target large language model.

[0018] In a possible implementation, after generating the final annotated data, the method further includes: constructing a knowledge graph of industrial and mining equipment according to the final annotated data; and sending the knowledge graph of industrial and mining equipment to a user interaction interface for data analysis and visualization operations.

[0019] In a second aspect, an embodiment of the present application provides a device for labeling device knowledge graph data, including:

[0020] A data management module, used to collect equipment characteristic data of multiple industrial and mining equipment, and perform data preprocessing on the equipment characteristic data to generate a data source with noise removed and a unified data format;

[0021] A labeling engine, used to obtain a labeling template, and input the labeling template and the data source into a pre-trained large language model for preliminary labeling processing, to generate preliminary labeling data corresponding to each of the plurality of industrial and mining equipment, wherein the preliminary labeling data includes entities, entity types and entity relationships corresponding to the plurality of industrial and mining equipment;

[0022] The annotation engine is further used to perform semi-automatic annotation processing on the entities, entity types and entity relationships corresponding to the multiple industrial and mining equipment based on the data source to generate final annotation data.

[0023] In a possible implementation, it also includes: a knowledge graph construction module, which is used to construct a knowledge graph of industrial and mining equipment based on the final labeled data; and a user interaction module, which is used to send the industrial and mining equipment knowledge graph to a user interaction interface for data analysis and visualization operations.

[0024] In a third aspect, an embodiment of the present application provides an electronic device, including: at least one processor and a memory;

[0025] The memory stores computer-executable instructions;

[0026] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method described in the first aspect above.

[0027] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer-executable instruction is stored. When a processor executes the computer-executable instruction, the device knowledge graph data labeling method described in the first aspect above is implemented.

[0028] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method for labeling device knowledge graph data as described in the first aspect.

[0029] The embodiment of the present application provides a method and device for labeling device knowledge graph data. Among them, the method first performs data preprocessing on the collected device characteristic data to generate a data source with noise removed and unified data format. Then the acquired standard template and data source are input into the pre-trained large language model for preliminary labeling processing to automatically generate preliminary labeling data. Finally, semi-automatic evaluation processing is performed based on the data source and the preliminary labeling data to generate the final labeling data. The process of manually analyzing and extracting data for building a device knowledge graph from massive device characteristic data is reduced, the automated labeling process shortens the data processing time, and when the labeling time increases, the automatic labeling process can ensure the consistency of the labeling process, avoid the subjectivity and inconsistency that may occur in manual labeling, and make the labeling accuracy higher. It not only saves labor costs and time costs, but also improves the labeling accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0031] Figure 1 A schematic diagram of an application scenario of the method for labeling device knowledge graph data provided in an embodiment of the present application;

[0032] Figure 2 A schematic diagram of a process for labeling device knowledge graph data provided in an embodiment of the present application;

[0033] Figure 3 A schematic diagram of data flow during the execution of the device knowledge graph data annotation method provided in an embodiment of the present application;

[0034] Figure 4 A schematic diagram of the structure of the device knowledge graph data annotation device provided in the embodiment of the present application Figure 1 ;

[0035] Figure 5 A schematic diagram of the structure of the device knowledge graph data annotation device provided in the embodiment of the present application Figure 2 ;

[0036] Figure 6 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application.

[0037] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0038] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0039] At present, the inventors have found that in the prior art, the efficient operation and maintenance and intelligent management of industrial and mining equipment lack unified and standard knowledge graph support, resulting in low data utilization and low accuracy of fault prediction and diagnosis. Therefore, the characteristic data of industrial and mining equipment can be annotated first, and then key feature information can be extracted from the data related to industrial and mining equipment to form a knowledge graph, so as to obtain a knowledge graph with clear structure and rich content, thereby significantly improving the efficiency and quality of data analysis and application.

[0040] To solve the above technical problems, the embodiments of the present application provide the following technical concepts for solving the problems: first, the collected device data is preprocessed to unify the data format, and then the preprocessed device data is preliminarily labeled using a large language model to improve the labeling efficiency and accuracy, and then the errors and omissions in the labeling are corrected through manual review and correction to further improve the labeling accuracy. The manually corrected data can also be used to continue training and iteratively optimize the large language model to improve the accuracy of subsequent labeling.

[0041] Figure 1 A schematic diagram of an application scenario of the method for labeling device knowledge graph data provided in an embodiment of the present application, such as Figure 1 As shown, it includes: a terminal 101 and a server 102.

[0042] Among them, the terminal 101 is used to respond to the user's interactive operation to obtain manually reviewed preliminary annotation data and correct the preliminary annotation data, display the final annotation data and the industrial and mining equipment knowledge map. The server 102 is used to receive industrial and mining equipment characteristic data transmitted from the cloud, other servers or mobile terminals, and perform data processing on the equipment characteristic data, use the large language model to obtain preliminary annotation data, and construct the industrial and mining equipment knowledge map.

[0043] Figure 2 This is a flow chart of the method for labeling device knowledge graph data provided in the embodiment of the present application. The execution subject of this embodiment can be Figure 1 The server 102 in the illustrated embodiment may also be other computer-related devices, and there is no particular limitation on this embodiment.

[0044] like Figure 2 As shown, the annotation method of the device knowledge graph data includes:

[0045] S201: Collect equipment characteristic data of a plurality of industrial and mining equipment, and perform data preprocessing on the equipment characteristic data to generate a data source with noise removed and a unified data format.

[0046] In this embodiment, the equipment characteristic data of industrial and mining equipment may be at least one of picture data, text data, audio data, and video data corresponding to the equipment related to industrial and mining. For example: product manuals, safety operating procedures, and other documents of conveyors in the media selection industry. Data preprocessing may be a process of summarizing and arranging equipment characteristic data and deleting data with low relevance to the equipment.

[0047] Specifically, in an optional embodiment of the present application, in step S201, the device characteristic data is preprocessed to generate a data source with noise removed and a unified data format, including:

[0048] S201a: Clean the device characteristic data to generate noise-removed data.

[0049] In this embodiment, the cleaning process may be a process of removing noise pixels in the image, deleting low-relevant data in the text, segmenting words, and removing stop words, etc., to increase the proportion of valid data.

[0050] S201b: Format the noise-removed data to generate data in a unified format as source data.

[0051] In this embodiment, formatting processing refers to converting data of different format types into data of the same format type, for example, converting TXT text, WORD text and PDF text in text data into text data in TXT or MARKDOWN format.

[0052] S202: Obtain a labeling template, and input the labeling template and the data source into a pre-trained large language model for preliminary labeling processing to generate preliminary labeling data corresponding to a plurality of industrial and mining equipment, wherein the preliminary labeling data includes entities, entity types, and entity relationships corresponding to the plurality of industrial and mining equipment.

[0053] In this embodiment, the annotation template may be a standardized template for annotation that is predefined according to the characteristics of industrial and mining equipment. The template may be a prompt word, equipment classification annotation, etc. In this embodiment, the pre-trained large language model may be a model obtained by combining a simple large language model with manual annotation using the equipment characteristic data related to industrial and mining equipment in advance. Among them, the simple large language model may be a data processing model with text recognition and relationship extraction functions.

[0054] Specifically, in an optional embodiment of the present application, obtaining the annotation template in step S202 includes:

[0055] S202a: Based on the pre-stored equipment characteristic data, define classification standards and different first entity categories for each type of industrial and mining equipment.

[0056] In this embodiment, the pre-stored equipment characteristic data may be previously stored data related to industrial and mining equipment. Defining the classification standard of each type of industrial and mining equipment may be a process of classifying equipment types according to the purpose, function, structure and other characteristics of each industrial and mining equipment. Defining different first entity categories may be a process of defining types according to the purpose of the equipment, for example: the first entity category may be a conveyor, a sieve and a shallow trough.

[0057] S202b: Define a first attribute field for each first entity category and determine a first relationship type between the first entity categories.

[0058] In this embodiment, the first attribute field may include text information related to the device attributes, such as the device's specification parameters, manufacturer information, operating conditions and environmental requirements. The first attribute field is intended to ensure the richness and descriptiveness of the data. The first relationship type may be a relationship between devices, such as: a component and a complete device, a device and operating safety regulations, a device and a maintenance process, etc.

[0059] S202c: Determine a labeling template according to the first entity category, the first attribute field, and the relationship type.

[0060] In this embodiment, the process of determining the annotation template can be for the equipment of an industry, defining multiple types of templates such as equipment type, equipment operation process safety, equipment maintenance, equipment failure, equipment inspection, etc. of the equipment in the industry.

[0061] In an optional embodiment of the present application, obtaining the annotation template in step S202 further includes:

[0062] S202d: Obtain annotation requirements regularly, and determine the second entity category according to the annotation requirements.

[0063] S202e: Define a second attribute field for the second entity category and determine a second relationship type between the second entity categories.

[0064] S202f: Determine a new annotation template according to the first entity category, the first attribute field, the first relationship type, the second entity type, the second attribute field, and the second relationship category.

[0065] In this embodiment, the periodicity can be set to a time related to the corresponding equipment working cycle, such as one month, one quarter, half a year, or one year. The labeling requirements can be obtained by leaving an extension interface when designing the labeling template. With the advancement of the technology used by the equipment and the development of the industry in which it is located, new entity categories and attribute fields are added to adapt to new labeling requirements. In order to achieve regular evaluation of the applicability of the template, ensure that these new labeling templates can cover the existing equipment types and industry requirements, and optimize them according to feedback. Make the final labeling data more accurate.

[0066] The above is the process of obtaining and updating the annotation template. The following will describe the use and training process of the pre-trained large language model in conjunction with the embodiments.

[0067] Based on the above embodiment, in an optional embodiment of the present application, in step S202, the data source is input into a pre-trained large language model to generate preliminary annotation data corresponding to multiple industrial and mining equipment, including: inputting the data source into the pre-trained large language model so that the pre-trained large language model automatically identifies the key information of the equipment from the data source, and classifies the key information of the equipment to obtain multiple entities and entity types corresponding to each entity, and performs relationship extraction on the data source to obtain entity relationships between different entities.

[0068] In this embodiment, the pre-trained large language model has named entity recognition and relationship extraction functions, and the data source is pre-processed. Therefore, when the data source is input into the pre-trained large language model, the pre-trained large language model automatically recognizes the key information of the equipment in the text information in the data source. These key information of the equipment may include equipment model, performance parameters, manufacturer, maintenance manual and fault handling method. The entity relationship between different entities obtained by the relationship extraction process can be the association relationship between different equipment, for example, the association relationship between "equipment type" and "maintenance content".

[0069] Based on the above embodiment, a method for labeling device knowledge graph data provided in an optional embodiment of the present application further includes:

[0070] Step A: Acquire original industrial and mining equipment data, and perform preprocessing operations on the original industrial and mining equipment data to obtain a training sample set and a verification sample set, wherein the training sample set includes multiple first industrial and mining equipment characteristic data, and the verification sample set includes multiple second industrial and mining equipment characteristic data.

[0071] In this embodiment, the original mining equipment characteristic data refers to the data related to the mining equipment that has been manually labeled and obtained manually labeled data. The first working condition equipment characteristic data and the second mining equipment characteristic data can be two parts of mining equipment characteristic data divided according to a specific ratio of the original working condition equipment characteristic data after preprocessing operations. For example: training sample set: verification sample set = 7:3.

[0072] Step B: training the initial large language model according to the first industrial and mining equipment characteristic data to obtain the large language model to be adjusted.

[0073] In this embodiment, the initial large language model can be a text recognition model or an image recognition model that can automatically perform OCR recognition. The large language model to be adjusted refers to a large language model obtained after training with a higher annotation accuracy than the initial large language model, and the large language model is suitable for the characteristics of the industrial and mining fields, and can more accurately understand the contextual information in the text related to industrial and mining equipment, so as to more accurately identify the entity types, attribute fields and relationship categories related to the equipment, etc., which are used to construct the triple data required for the equipment knowledge graph.

[0074] Step C: Determine the labeling accuracy of the large language model to be adjusted through multiple second device characteristic data. If the labeling accuracy is less than the first preset threshold, perform a first adjustment operation on the large language model to be adjusted until the labeling accuracy is greater than or equal to the first preset threshold, thereby obtaining a pre-trained large language model.

[0075] In this embodiment, the second device characteristic data may be data used to verify the labeling accuracy of the large language model to be adjusted, so as to achieve the purpose of optimizing the correlation coefficient of the large language model to be adjusted. The first adjustment operation refers to the process of iteratively optimizing the large language model to be adjusted according to the verification sample set. Until the labeling accuracy of the large language model to be adjusted after the iteration is greater than or equal to the first preset threshold. The first preset threshold may be a preset value, for example, the first preset threshold may be 90%. At this time, the large language model to be adjusted obtained in the last iteration may be determined as the pre-trained large language model.

[0076] Based on the above embodiment, in an optional embodiment of the present application, in order to increase the amount of data for training the large language model and improve the accuracy of the final pre-trained large language model, after obtaining the original mining equipment data in step A in the above embodiment, it also includes:

[0077] Step D: performing data augmentation processing on the original industrial and mining equipment data through the initial large language model to obtain augmented industrial and mining equipment data, wherein the augmented industrial and mining equipment data includes the original industrial and mining equipment data and the newly added industrial and mining equipment data.

[0078] In this embodiment, data augmentation processing refers to the process of using the initial large language model to expand the original industrial and mining equipment data to increase the amount of data. For example, the original industrial and mining equipment data originally extracted contains centrifuge fault data that only includes high discharge moisture and pump fault data. After data augmentation processing, a set of centrifuge screen basket breakage fault data will be added. After obtaining the augmented industrial and mining equipment data, the subsequent model training steps can be continued. More data is used for model training and optimization, which can make the annotation accuracy of the final pre-trained large language model higher.

[0079] Based on the above embodiment, in order to further improve the annotation accuracy of the pre-trained large language model, the annotation method of the device knowledge graph data provided as an optional embodiment of the present application also includes:

[0080] Step E: Determine the labeling accuracy of the pre-trained large language model through the final labeled data. If the labeling accuracy is less than the second preset threshold, perform a second adjustment operation on the pre-trained large language model until the labeling accuracy is greater than or equal to the second preset threshold, thereby obtaining the target large language model.

[0081] In this embodiment, the second preset threshold is a value greater than the first preset threshold, for example, if the first preset threshold is 90%, the second preset threshold may be 95%. The second adjustment operation refers to the process of adjusting the relevant parameters of the pre-trained large language model, for example, the second adjustment operation may be LoRa fine-tuning. The target large language model refers to the large language model obtained after model iterative optimization.

[0082] Figure 3 A schematic diagram of data flow during the execution of the device knowledge graph data annotation method provided in an embodiment of the present application.

[0083] like Figure 3 As shown, the final annotated data obtained by the annotation process on the right can be used as supplementary data for the model training and optimization process on the left. The process of optimizing the pre-trained large language model using the final annotated data is similar to the process of using multiple second device characteristic data and subsequent processes in step C of the above embodiment, so this embodiment will not be repeated here.

[0084] like Figure 3As shown, for the collected equipment characteristic data of industrial and mining equipment, a data source is generated after preprocessing. The data source and annotation template are input into the pre-trained large language model to generate a preliminary annotation result. For the preliminary annotation result, on the one hand, the triples for equipment knowledge graph data that may have errors or problems in the preliminary annotation result can be quickly found through automatic evaluation. Then, the preliminary annotation result can be manually checked, and the final annotation data can be obtained through manual evaluation combined with automatic evaluation. The final annotation data can be fed back to the training and optimization of the pre-trained large language model. This makes the annotation accuracy of the subsequent pre-trained large language model higher and higher.

[0085] S203: Based on the data source, semi-automatic evaluation and processing is performed on entities, entity types and entity relationships corresponding to multiple industrial and mining equipment to generate final labeled data.

[0086] In this embodiment, the semi-automatic evaluation process can be an audit process in response to the results of the manual reference automatic audit to check for possible omissions or errors in the preliminary annotation results. Audit users in the mining equipment industry use their own annotation experience to review the preliminary annotation results and correct the incorrect or omitted annotations to ensure the accuracy of the final annotation data.

[0087] The above embodiment is intended to describe the implementation process of the data annotation method of the equipment knowledge graph. The ultimate purpose of data annotation is to extract key feature information from the data related to industrial and mining equipment to obtain the triple data (corresponding to the final annotation data in the above embodiment) that is ultimately used to construct the equipment knowledge graph. The equipment knowledge graph has a clear structure and rich content, which can significantly improve the efficiency and quality of data analysis and application. Based on this, in an optional embodiment of the present application, after the final annotation data is generated in step S203, it also includes:

[0088] Step F: Construct the knowledge graph of industrial and mining equipment based on the final labeled data.

[0089] In this embodiment, the process of constructing the knowledge graph of industrial and mining equipment is a process of using the existing equipment characteristic knowledge graph schema and the final annotation data to construct a knowledge graph, wherein the schema includes relationships, entities and attributes.

[0090] Step G: Send the industrial and mining equipment knowledge graph to the user interaction interface for data analysis and visualization operations.

[0091] In this embodiment, the user interaction interface is intended to provide a responsive interface for user interaction. The user interaction interface supports users to view and edit knowledge graphs, as well as perform data analysis and visualization.

[0092] In summary, the method for labeling device knowledge graph data provided in the embodiment of the present application first performs data preprocessing on the collected device characteristic data to generate a data source with noise removed and unified data format. Then the acquired standard template and data source are input into the pre-trained large language model for preliminary labeling processing to automatically generate preliminary labeling data. Finally, semi-automatic evaluation processing is performed based on the data source and the preliminary labeling data to generate the final labeling data. The process of manually analyzing and extracting data for building a device knowledge graph from massive device characteristic data is reduced, the automated labeling process shortens the data processing time, and when the labeling time increases, the automatic labeling process can ensure the consistency of the labeling process, avoid the subjectivity and inconsistency that may occur in manual labeling, and make the labeling accuracy higher. It not only saves labor costs and time costs, but also improves the labeling accuracy.

[0093] At the same time, a large amount of original industrial and mining equipment data is used for model training and model optimization to obtain a pre-trained large language model that is more suitable for the characteristics of the industrial and mining field. In the training process, data augmentation is performed to further increase the data volume to improve the annotation accuracy of the final pre-trained large language model.

[0094] At the same time, by designing annotation templates and updating data annotation templates as needed, this application can support the annotation needs of more types of industrial and mining equipment and improve scalability.

[0095] At the same time, the final labeled data is used to iteratively optimize the pre-trained large language model, so that the pre-trained large language model can continuously improve the automatic labeling efficiency and labeling accuracy through continuous learning and optimization.

[0096] At the same time, the final labeled data is used to build a device knowledge graph for users to view and analyze, thereby significantly improving the efficiency and quality of data analysis and application.

[0097] Figure 4 A schematic diagram of the structure of the device knowledge graph data annotation device provided in the embodiment of the present application Figure 1 The device includes: a data management module 41 and a labeling engine 42.

[0098] The data management module 41 is used to collect equipment characteristic data of a plurality of industrial and mining equipment, and perform data preprocessing on the equipment characteristic data to generate a data source with noise removed and a unified data format.

[0099] The labeling engine 42 is used to obtain a labeling template, and input the labeling template and the data source into a pre-trained large language model for preliminary labeling processing to generate preliminary labeling data corresponding to each of the multiple industrial and mining equipment, wherein the preliminary labeling data includes entities, entity types and entity relationships corresponding to the multiple industrial and mining equipment.

[0100] The annotation engine 42 is also used to perform semi-automatic annotation processing on entities, entity types and entity relationships corresponding to multiple industrial and mining equipment based on the data source to generate final annotation data.

[0101] In an optional embodiment of the present application, the data management module 41 is specifically used to: clean the device characteristic data to generate noise-removed data; format the noise-removed data to generate unified format data as source data.

[0102] In an optional embodiment of the present application, the labeling engine 42 is specifically used to: define the classification standards and different first entity categories for each type of industrial and mining equipment based on pre-stored equipment characteristic data; define the first attribute field for each first entity category and determine the first relationship type between the first entity categories; determine the labeling template according to the first entity category, the first attribute field and the relationship type.

[0103] In an optional embodiment of the present application, the labeling engine 42 is specifically used to: input the data source into the pre-trained large language model so that the pre-trained large language model automatically identifies the key information of the device from the data source, and classifies the key information of the device to obtain multiple entities and entity types corresponding to each entity, and performs relationship extraction on the data source to obtain entity relationships between different entities.

[0104] In an optional embodiment of the present application, the annotation engine 42 is also used to: periodically obtain annotation requirements and determine the second entity category based on the annotation requirements; define second attribute fields for the second entity category and determine the second relationship type between each second entity category; determine a new annotation template based on the first entity category, the first attribute field, the first relationship type, the second entity type, the second attribute field and the second relationship category.

[0105] In an optional embodiment of the present application, the data management module 41 is further used to: obtain original industrial and mining equipment data, and perform preprocessing operations on the original industrial and mining equipment data to obtain a training sample set and a verification sample set, wherein the training sample set includes multiple first industrial and mining equipment characteristic data, and the verification sample set includes multiple second industrial and mining equipment characteristic data; train the initial large language model according to the first industrial and mining equipment characteristic data to obtain the large language model to be adjusted; determine the labeling accuracy of the large language model to be adjusted through multiple second equipment characteristic data, and if the labeling accuracy is less than a first preset threshold, perform a first adjustment operation on the large language model to be adjusted until the labeling accuracy is greater than or equal to the first preset threshold, and obtain a pre-trained large language model.

[0106] In an optional embodiment of the present application, the data management module 41 is further used to: perform data augmentation processing on the original industrial and mining equipment data through the initial large language model to obtain augmented industrial and mining equipment data, wherein the augmented industrial and mining equipment data includes the original industrial and mining equipment data and the newly added industrial and mining equipment data.

[0107] In an optional embodiment of the present application, the data management module 41 is also used to: determine the labeling accuracy of the pre-trained large language model through the final labeled data; if the labeling accuracy is less than a second preset threshold, perform a second adjustment operation on the pre-trained large language model until the labeling accuracy is greater than or equal to the second preset threshold, thereby obtaining a target large language model.

[0108] Figure 5 A schematic diagram of the structure of the device knowledge graph data annotation device provided in the embodiment of the present application Figure 2 In an optional embodiment of the present application, the device for labeling device knowledge graph data also includes: a knowledge graph construction module 43 and a user interaction module 44.

[0109] The knowledge graph construction module 43 is used to construct a knowledge graph of industrial and mining equipment based on the final labeled data;

[0110] The user interaction module 44 is used to send the industrial and mining equipment knowledge graph to the user interaction interface for data analysis and visualization operations.

[0111] The device knowledge graph data annotation device provided in this embodiment can be used to execute the technical solution of the above-mentioned method embodiment. Its implementation principle and technical effects are similar, and this embodiment will not be repeated here.

[0112] Figure 6 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, the electronic device includes: at least one processor 601 and a memory 602 .

[0113] The memory 602 is used to store computer-executable instructions.

[0114] The processor 601 is used to execute the computer-executable instructions stored in the memory 602 to implement the various steps involved in the above method embodiment. For details, please refer to the relevant description in the above method embodiment.

[0115] Optionally, the memory 602 may be independent or integrated with the processor 601 .

[0116] When the memory 602 is independently provided, the electronic device further includes a bus 603 for connecting the memory 602 and the processor 601 .

[0117] An embodiment of the present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above-mentioned device knowledge graph data labeling device method is implemented.

[0118] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned device knowledge graph data labeling method.

[0119] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0120] The modules described above as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to implement the solution of this embodiment.

[0121] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each module may exist physically separately, or two or more modules may be integrated into one unit. The unit composed of the above modules may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0122] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the method of each embodiment of the present application.

[0123] It should be understood that the above processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the invention can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor.

[0124] The memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0125] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0126] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0127] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic device or a main control device.

[0128] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0129] The present invention is intended to illustrate the technical solution of the present application rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solution from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for labeling device knowledge graph data, characterized in that: include: Collecting equipment characteristic data of multiple industrial and mining equipment, and performing data preprocessing on the equipment characteristic data to generate a data source with noise removed and a unified data format; Obtaining a labeling template, and inputting the labeling template and the data source into a pre-trained large language model for preliminary labeling processing, to generate preliminary labeling data corresponding to each of the plurality of industrial and mining equipment, wherein the preliminary labeling data includes entities, entity types, and entity relationships corresponding to the plurality of industrial and mining equipment; Based on the data source, semi-automatic evaluation and processing are performed on the entities, entity types and entity relationships corresponding to the plurality of industrial and mining equipment to generate final labeled data.

2. The method according to claim 1, characterized in that The preprocessing of the device characteristic data to generate a data source with noise removed and a unified data format includes: Cleaning the device characteristic data to generate noise-removed data; The noise-removed data is formatted to generate data in a unified format as source data.

3. The method according to claim 1, characterized in that The step of obtaining the annotation template includes: Based on the pre-stored equipment characteristic data, define the classification standard and different first entity categories of each type of industrial and mining equipment; defining a first attribute field for each first entity category and determining a first relationship type between the first entity categories; A labeling template is determined according to the first entity category, the first attribute field and the relationship type.

4. The method according to claim 1, characterized in that: The step of inputting the data source into a pre-trained large language model to generate preliminary labeling data corresponding to each of the plurality of industrial and mining equipment includes: The data source is input into a pre-trained large language model so that the pre-trained large language model automatically identifies key device information from the data source, classifies the key device information to obtain multiple entities and entity types corresponding to each entity, and performs relationship extraction on the data source to obtain entity relationships between different entities.

5. The method according to claim 3, characterized in that: The obtaining of the annotation template further includes: Regularly obtaining annotation requirements, and determining a second entity category according to the annotation requirements; defining a second attribute field for the second entity category and determining a second relationship type between the second entity categories; A new annotation template is determined according to the first entity category, the first attribute field, the first relationship type, the second entity type, the second attribute field and the second relationship category.

6. The method according to claim 1, characterized in that Also includes: Acquire original mining equipment data, and perform preprocessing operations on the original mining equipment data to obtain a training sample set and a verification sample set, wherein the training sample set includes a plurality of first mining equipment characteristic data, and the verification sample set includes a plurality of second mining equipment characteristic data; According to the first industrial and mining equipment characteristic data, an initial large language model is trained to obtain a large language model to be adjusted; The labeling accuracy of the large language model to be adjusted is determined through multiple second device characteristic data. If the labeling accuracy is less than a first preset threshold, a first adjustment operation is performed on the large language model to be adjusted until the labeling accuracy is greater than or equal to the first preset threshold, so as to obtain a pre-trained large language model.

7. The method according to claim 6, characterized in that After obtaining the original industrial and mining equipment data, the method further includes: The original industrial and mining equipment data is subjected to data augmentation processing by using the initial large language model to obtain augmented industrial and mining equipment data, wherein the augmented industrial and mining equipment data includes the original industrial and mining equipment data and the newly added industrial and mining equipment data.

8. The method according to claim 7, characterized in that Also includes: The labeling accuracy of the pre-trained large language model is determined by the final labeled data. If the labeling accuracy is less than a second preset threshold, a second adjustment operation is performed on the pre-trained large language model until the labeling accuracy is greater than or equal to the second preset threshold, thereby obtaining a target large language model.

9. The method according to any one of claims 1 to 8, characterized in that After generating the final annotation data, the method further includes: Constructing a knowledge graph of industrial and mining equipment according to the final labeled data; The industrial and mining equipment knowledge graph is sent to a user interaction interface for data analysis and visualization operations.

10. A device for labeling equipment knowledge graph data, characterized in that: include: A data management module, used to collect equipment characteristic data of multiple industrial and mining equipment, and perform data preprocessing on the equipment characteristic data to generate a data source with noise removed and a unified data format; A labeling engine, used to obtain a labeling template, and input the labeling template and the data source into a pre-trained large language model for preliminary labeling processing, to generate preliminary labeling data corresponding to each of the plurality of industrial and mining equipment, wherein the preliminary labeling data includes entities, entity types and entity relationships corresponding to the plurality of industrial and mining equipment; The annotation engine is further used to perform semi-automatic annotation processing on the entities, entity types and entity relationships corresponding to the multiple industrial and mining equipment based on the data source to generate final annotation data.

Images